The Timed Up and Go Test does not predict length of stay on an acute geriatric ward

This study aimed to determine whether the admission Timed Up and Go Test (TUG) predicted the length of stay of patients in an acute geriatric ward. Consecutive patients were quasi-randomly allocated to either a model development sample or a model validation sample. Multivariate Cox proportional hazards regression was used to model length of stay. Variables considered for inclusion in the development model were risk factors for length of stay reported in the literature and univariate predictors from our dataset (p < 0.05). Variables selected for use in the development sample were then tested in the validation sample. Of 2463 patients of mean age 82.1 years, 932 (37.8%) were able to complete the TUG. Despite a significant, though weak, relationship between the length of stay and the TUG time (Spearman coefficient 0.18, p < 0.001), no time clearly identified patients with longer length of stay. Patients unable to complete the TUG had a median length of stay of 11 days (IQR 7 to 18), 40% longer than those able to complete the TUG (median 8 days, IQR 8 to 12, p < 0.001). Other significant (p < 0.05) predictors of length of stay in both samples were number of active medical diagnoses, referral from the emergency department, in-patient fall, and diagnosis of ulcer or infection. The admission TUG time should not be used to screen for patients likely to have longer lengths of stay. The value of the TUG lies in determining the patient's ability to complete it, rather than the time taken.